A robot intelligence evaluation method and system

Abstract

The present invention discloses a robot intelligence evaluation method and system, the method comprising the following steps: constructing an intelligence evaluation index system based on the perception, judgment, decision-making and execution of the robot; optimizing the training intelligence evaluation index, and obtaining the weights of various indicators for different types of robots; constructing a system testing process, and selecting intelligence evaluation indicators to evaluate the intelligence of the robot; the advantages of the present invention are: comprehensively evaluating the intelligence level of the robot in multiple dimensions such as perception, judgment, decision-making and execution.

Description

Technical Field

[0001] The present invention relates to the field of robot intelligence evaluation, and in particular to a robot intelligence evaluation method and system. Background Art

[0002] In the field of robot intelligence evaluation, evaluating the intelligence of a robot is to obtain its intelligence level in different tasks and scenarios. At present, the mainstream evaluation methods are mainly focused on behavioral testing, that is, evaluating the robot's ability by observing the robot's behavioral performance in a specific task or scenario. For example, Chinese Patent Publication No. CN116901135A discloses a system and method for evaluating the perception ability of a substation inspection robot. The system includes: a state detection device for detecting the state of the inspection robot; a recognition ability detection device for detecting the recognition ability of the inspection robot; a communication detection device for detecting the communication status of the inspection robot; and a control device for controlling the state detection device, the recognition ability detection device and the communication detection device to detect the inspection robot. This patent application can perform multiple ability tests on the inspection robot without the need to perform different ability tests separately through multiple detection devices, thus achieving modularization, integration and simple operation. However, the above method focuses on behavioral testing, only focusing on the accuracy of the robot's behavior, and there is no evaluation method related to the robot's intelligence level. Therefore, it is impossible to comprehensively evaluate the robot's intelligence level in multiple dimensions such as perception, judgment, decision-making and execution. Summary of the invention

[0003] The technical problem to be solved by the present invention is how to comprehensively evaluate the intelligence level of a robot in multiple dimensions such as perception, judgment, decision-making and execution.

[0004] The present invention solves the above technical problems through the following technical means: a robot intelligence evaluation method, comprising the following steps:

[0005] Step 1: Build an intelligence evaluation index system based on the robot's perception, judgment, decision-making and execution;

[0006] Step 2: Optimize the training intelligence evaluation indicators and obtain the weights of various indicators for different types of robots;

[0007] Step 3: Build a system testing process and select intelligence evaluation indicators to evaluate the intelligence of the robot.

[0008] Furthermore, the step one includes: the intelligence evaluation index system includes a control steady-state performance index, an obstacle avoidance sensitivity index, a vehicle body stability index, a target positioning accuracy index and a planning performance index, the control steady-state performance index and the vehicle body stability index are used to evaluate the robot's performance in execution ability; the obstacle avoidance sensitivity index is used to evaluate the robot's performance in judgment ability; the target positioning accuracy index is used to evaluate the robot's performance in perception ability; the planning performance index is used to evaluate the robot's performance in decision-making and planning ability.

[0009] Furthermore, the control steady-state performance indicators include average vehicle speed, average left speed, average right speed and average turning angle; the obstacle avoidance sensitivity indicators include average vehicle speed, unit absolute trajectory error and unit relative trajectory error; the vehicle stability indicators include average turning angle; the target positioning accuracy indicators include unit absolute trajectory error and unit relative trajectory error; the planning performance indicators include path time, path distance, planned path smoothness, average turning radius and maximum turning angle.

[0010] Furthermore, the step 2 includes:

[0011] Step 201, constructing a data set;

[0012] Step 202: for the intelligence evaluation, construct an objective function by accumulating various indicators and corresponding indicator data;

[0013] Step 203: Solve the objective function to obtain the weights of various indicators of the evaluation tasks for different types of robots.

[0014] Furthermore, the step 201 includes:

[0015] First, a specific scenario task is constructed, and a robot of a specified type performs the task multiple times to obtain data indicators of the motion process. Assuming there are n indicators, the task is performed t times, and the tasks performed by the robot are scored at the same time. The data indicators and task scores are combined to obtain a matrix with t rows and (n+1) columns.

[0016] Furthermore, the step 202 includes:

[0017] The objective function is constructed as F = W1 X1 + W2X2 + ... + W n X n

[0018] Among them, F is the scoring result of task execution, W1,W2,...,W n is the weight of each indicator, which is a variable; X1, X2, ..., X n is the corresponding data indicator;

[0019] The constraints of the objective function are: W1+W2+...+W n =1;W i >=0,W i <=1,i=1,2,...,n.

[0020] Furthermore, the step 203 includes:

[0021] For different tasks, the ultimate goal of the scoring result F of task execution is to obtain the minimum scoring result or the maximum scoring result. The objective function is solved by the gradient descent method. Through the iterative optimization process, the objective function F is maximized or minimized to obtain the optimal weight, thereby obtaining the weight of each indicator.

[0022] Furthermore, the system testing process in step 3 includes:

[0023] Step 301, determine the test object: enter the test platform, and select the robot component or functional module to be tested in the platform;

[0024] Step 302, select intelligent evaluation indicators: select an intelligent evaluation indicator system according to the specific task or scenario of the robot;

[0025] Step 303, adding test tasks and parameters: adding test tasks and setting corresponding test parameters;

[0026] Step 304, obtaining simulation or real data: starting the simulation system for testing, the simulation system performs simulation according to the set test tasks and parameters, and generates simulation data after the simulation is completed or imports motion data collected in the real environment for evaluation;

[0027] Step 305, data processing and evaluation report: obtain simulation or real data and save it in the database, clean and denoise the data, weight the simulation results or real data according to the weights of different indicators to obtain evaluation scores, perform evaluation, and generate an intelligent evaluation report.

[0028] The present invention provides a robot intelligence evaluation system, comprising:

[0029] The indicator system building module is used to build an intelligence evaluation indicator system based on the robot's perception, judgment, decision-making and execution;

[0030] The indicator system training module is used to optimize the training intelligence evaluation indicators and obtain the weights of various indicators for different types of robots;

[0031] The intelligence evaluation module is used to build a system testing process and select intelligence evaluation indicators to evaluate the intelligence of the robot.

[0032] Furthermore, the index system construction module is also used for: the intelligence evaluation index system includes a control steady-state performance index, an obstacle avoidance sensitivity index, a vehicle body stability index, a target positioning accuracy index and a planning performance index. The control steady-state performance index and the vehicle body stability index are used to evaluate the robot's performance in execution ability; the obstacle avoidance sensitivity index is used to evaluate the robot's performance in judgment ability; the target positioning accuracy index is used to evaluate the robot's performance in perception ability; and the planning performance index is used to evaluate the robot's performance in decision-making and planning ability.

[0033] Furthermore, the control steady-state performance indicators include average vehicle speed, average left speed, average right speed and average turning angle; the obstacle avoidance sensitivity indicators include average vehicle speed, unit absolute trajectory error and unit relative trajectory error; the vehicle stability indicators include average turning angle; the target positioning accuracy indicators include unit absolute trajectory error and unit relative trajectory error; the planning performance indicators include path time, path distance, planned path smoothness, average turning radius and maximum turning angle.

[0034] Furthermore, the indicator system training module is also used for:

[0035] Step 201, constructing a data set;

[0036] Step 202: for the intelligence evaluation, construct an objective function by accumulating various indicators and corresponding indicator data;

[0037] Step 203: Solve the objective function to obtain the weights of various indicators of the evaluation tasks for different types of robots.

[0038] Furthermore, the step 201 includes:

[0039] First, a specific scenario task is constructed, and a robot of a specified type performs the task multiple times to obtain data indicators of the motion process. Assuming there are n indicators, the task is performed t times, and the tasks performed by the robot are scored at the same time. The data indicators and task scores are combined to obtain a matrix with t rows and (n+1) columns.

[0040] Furthermore, the step 202 includes:

[0041] The objective function is constructed as F = W1 X1 + W2X2 + ... + W n X n

[0042] Among them, F is the scoring result of task execution, W1,W2,...,W n is the weight of each indicator, which is a variable; X1, X2, ..., X n is the corresponding data indicator;

[0043] The constraints of the objective function are: W1+W2+...+W n =1;W i >=0,W i <=1,i=1,2,...,n.

[0044] Furthermore, the step 203 includes:

[0045] For different tasks, the ultimate goal of the scoring result F of task execution is to obtain the minimum scoring result or the maximum scoring result. The objective function is solved by the gradient descent method. Through the iterative optimization process, the objective function F is maximized or minimized to obtain the optimal weight, thereby obtaining the weight of each indicator.

[0046] Furthermore, the system testing process in the intelligence evaluation module includes:

[0047] Step 301, determine the test object: enter the test platform, and select the robot component or functional module to be tested in the platform;

[0048] Step 302, select intelligent evaluation indicators: select an intelligent evaluation indicator system according to the specific task or scenario of the robot;

[0049] Step 303, adding test tasks and parameters: adding test tasks and setting corresponding test parameters;

[0050] Step 304, obtaining simulation or real data: starting the simulation system for testing, the simulation system performs simulation according to the set test tasks and parameters, and generates simulation data after the simulation is completed or imports motion data collected in the real environment for evaluation;

[0051] Step 305, data processing and evaluation report: obtain simulation or real data and save it in the database, clean and denoise the data, weight the simulation results or real data according to the weights of different indicators to obtain evaluation scores, perform evaluation, and generate an intelligent evaluation report.

[0052] The advantages of the present invention are:

[0053] (1) The present invention first constructs a robot intelligence evaluation index system based on the robot's perception, judgment, decision-making and execution; then, during the intelligence evaluation process, the weights of the intelligence evaluation indicators can be optimized according to different types of robots; finally, an intelligent operation process is provided for robot intelligence evaluation to comprehensively evaluate the robot's intelligence level in multiple dimensions such as perception, judgment, decision-making and execution.

[0054] (2) The present invention relates to the field of robot intelligence assessment. Through three-dimensional intelligent methods, namely, the construction of an indicator system for assessing intelligence, the optimization of training assessment parameters, and the intelligent assessment process, the intelligent assessment method is combined with the kinematic indicators of the robot's task execution to achieve a comprehensive and accurate assessment of the robot's intelligence level and effectively output the assessment results. These dimensional intelligent methods ensure the comprehensiveness, accuracy, and reliability of the assessment, and can comprehensively and accurately assess the robot's intelligence level and effectively output the assessment results, providing a well-founded reference for robot research and development and application. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 A flowchart of a robot intelligence evaluation method disclosed in an embodiment of the present invention;

[0056] Figure 2 A schematic diagram of a system test flow in a robot intelligence evaluation method disclosed in an embodiment of the present invention;

[0057] Figure 3 This is a diagram of an interface of tracked robot evaluation results in a robot intelligence evaluation method disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in combination with the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] Example 1

[0060] like Figure 1 As shown, the present invention provides a robot intelligence evaluation method, comprising the following steps:

[0061] S1. Build an intelligence evaluation index system based on the robot's perception, judgment, decision-making and execution;

[0062] In the field of human intelligence research, human behavior is to obtain information from the outside world, then conduct cognition and decision-making, and finally transform it into behavior. This process can be divided into four key steps: obtaining information, processing and thinking, judgment, and execution, to reflect the basic characteristics of human intelligence. First, humans obtain various information from the environment through sensory systems (such as vision, hearing, etc.). Then, humans recognize, process and think about this information, integrating and analyzing different viewpoints, concepts and experiences. Next, humans make judgments, evaluate feasible options, and choose the best course of action. Finally, based on judgments, humans perform corresponding behaviors to achieve specific goals or complete tasks. This four-step process reflects the basic way human intelligence processes information and makes decisions.

[0063] When constructing a robot intelligence evaluation index system, we can draw on the human thinking and behavior process to evaluate and measure the robot's intelligence level. Therefore, we can establish an index system based on intelligent perception, judgment, decision-making and execution to evaluate the robot's intelligence.

[0064] In terms of robot intelligence evaluation, an evaluation index system is constructed by collecting the robot's motion data when performing specific tasks or scenarios. This index system is based on the four aspects of robot intelligence: perception, judgment, decision-making and execution. Specifically, the intelligence index is constructed from the following perspectives: the target positioning accuracy index is used to evaluate the robot's performance in perception ability; the obstacle avoidance sensitivity index is used to evaluate the robot's performance in judgment ability; the planning performance index is used to evaluate the robot's performance in planning ability; the control steady-state performance and vehicle body stability index are used to evaluate the robot's performance in execution ability. These indicators will be constructed based on the robot's motion data. The intelligence index is decomposed into kinematic indicators based on the physical quantities of the robot's motion process. The following Table 1 is a specific indicator classification.

[0065] Table 1 Robot intelligence evaluation index system

[0066]

[0067]

[0068] Through such an evaluation index system, the robot's performance in four aspects, namely perception, judgment, decision-making and execution, can be quantified and evaluated based on the robot's motion data, and ultimately a comprehensive intelligence evaluation of the robot can be completed from these four aspects.

[0069] S2. Optimize the training intelligence evaluation index and obtain the weights of various indicators for different types of robots. In order to intelligently evaluate the intelligence of different types of robots, this embodiment uses a machine learning algorithm to calculate the weights of various indicators. The specific steps are as follows:

[0070] Step 201, constructing a data set: First, construct a specific scenario task, and have a robot of a specified type perform the task multiple times, obtain data indicators of the motion process, assume that there are n indicators, perform t tasks, and score the tasks performed by the robot at the same time to evaluate the motion stability and execution accuracy of the task. Combine the data indicators and task scores to obtain a matrix with t rows and (n+1) columns.

[0071] Step 202: for the intelligent evaluation, the objective function is constructed by accumulating various indicators and corresponding indicator data; the constructed objective function is F=W1X1+W2X2+...+W n X n

[0072] Among them, F is the scoring result of task execution, W1,W2,...,W n is the weight of each indicator, which is a variable; X1, X2, ..., X n is the corresponding data indicator;

[0073] The constraints of the objective function are: W1+W2+...+W n =1;W i >=0,W i <=1,i=1,2,...,n.

[0074] Step 203, solving the objective function: for different tasks, the ultimate goal of the scoring result F of task execution is to obtain the minimum scoring result or the maximum scoring result. The gradient descent method or other optimization algorithms can be used to solve the objective function. Through the iterative optimization process, the objective function F is maximized or minimized to obtain the optimal weight, thereby obtaining the weight of each indicator.

[0075] The above optimization training of intelligent evaluation indicators is used to calculate the weight of each indicator, and then the importance of different indicators is quantified according to the weight. This weight calculation process is based on machine learning algorithms, making the evaluation results more accurate and reliable.

[0076] S3. Construct a system test process and select intelligence evaluation indicators to evaluate the intelligence of the robot. In this embodiment, in order to improve the accuracy of the robot intelligence test and reduce manpower and time costs, a set of intelligent system test processes is constructed to more accurately and efficiently evaluate the performance of the robot. Figure 2 As shown, the specific process is as follows:

[0077] Step 301, determine the test object: the user logs into the system and enters the test platform. In the platform, the robot component or functional module to be tested can be selected.

[0078] Step 302: Select intelligent evaluation indicators: The user can select an intelligent evaluation indicator system according to the specific tasks or scenarios of the robot. The indicator system may include indicators in terms of perception, judgment, decision-making and execution.

[0079] Step 303, adding test tasks and parameters: In the test task module, the user can add test tasks and set corresponding test parameters. When adding test parameters, different test scenario information can be selected to ensure that the test covers a variety of situations.

[0080] Step 304, obtain simulation or real data: The user can click the start button on the interface to start the simulation system for testing. The simulation system will simulate according to the set test tasks and parameters, and generate simulation data after the simulation is completed. In addition, the system also supports importing motion data collected in the real environment for evaluation.

[0081] Step 305, data processing and evaluation report: The evaluation system will obtain simulated or real data and save it to the database. The user can select the data preprocessing method to clean and reduce the noise of the data. Finally, the system will evaluate the simulation results and generate an intelligent evaluation report, providing detailed performance evaluation results and analysis.

[0082] The above intelligent testing process can more accurately evaluate the intelligent performance of the robot, while reducing manpower and time costs and improving testing efficiency.

[0083] In a specific scenario task, the operation steps of the intelligent testing process are as follows:

[0084] (1) Index construction: Construct an evaluation index system based on the components or functions of the robot to be tested. Indicators can be constructed by system recommendation or self-selection.

[0085] (2) Test task issuance: Create a test task, including detailed information such as data batch, creation time, data type, test environment, indicator information, test scenario, test purpose, and evaluation criteria. This step ensures the accuracy and completeness of the test task.

[0086] (3) Case generation: After the test task is created, click on the case generation button and the system will generate the test case. The test case includes basic information, test index information, and print case information. At the same time, click the button to start the simulation system interface, perform simulation testing, and generate simulation data after the simulation is completed.

[0087] (4) Data acquisition: Enter the data acquisition module and save the simulation data into the database for subsequent data processing and evaluation.

[0088] (5) Data cleaning: In the data cleaning step, users can select the required compensation algorithm, noise reduction algorithm, smoothness and other parameters. After clicking the start button, the system will automatically clean the data. The cleaning results are dynamically displayed in the list through a curve graph, and the data cleaning progress bar will be highlighted as the cleaning progresses.

[0089] (6) Intelligent evaluation: Enter the intelligent evaluation module and select the required intelligent evaluation algorithm. After the evaluation is completed, the system will analyze the intelligence score and evaluation level of the simulation data.

[0090] Through the above optimized process, the intelligent performance of the robot can be accurately evaluated in specific scenario tasks, and detailed evaluation results and reports can be provided. This process ensures the standardization of evaluation and the accuracy of operation, which helps to improve the reliability and efficiency of evaluation. Figure 3 The following is a diagram of the evaluation results of the tracked robot. Based on the intelligent evaluation results, the robot's intelligence level in different task scenarios is output. The evaluation results are presented in the form of numerical values, charts, and text, which are used to compare the performance of different robots and guide the improvement and optimization of robots.

[0091] Through the above technical scheme, the present invention adopts three-dimensional intelligent methods, namely, the construction of an indicator system for evaluating intelligence, the optimization of training evaluation parameters and the intelligent evaluation process. The intelligent evaluation method is combined with the kinematic indicators of the robot's task execution to achieve a comprehensive and accurate evaluation of the robot's intelligence level and effectively output the evaluation results. These dimensional intelligent methods ensure the comprehensiveness, accuracy and reliability of the evaluation, can comprehensively and accurately evaluate the robot's intelligence level, and effectively output the evaluation results, providing a well-founded reference for robot research and development and application.

[0092] Example 2

[0093] Based on Example 1, Example 2 of the present invention further provides a robot intelligence evaluation system, including:

[0094] The indicator system building module is used to build an intelligence evaluation indicator system based on the robot's perception, judgment, decision-making and execution;

[0095] The indicator system training module is used to optimize the training intelligence evaluation indicators and obtain the weights of various indicators for different types of robots;

[0096] The intelligence evaluation module is used to build a system testing process and select intelligence evaluation indicators to evaluate the intelligence of the robot.

[0097] Specifically, the index system construction module is also used for: the intelligence evaluation index system includes control steady-state performance indicators, obstacle avoidance sensitivity indicators, vehicle stability indicators, target positioning accuracy indicators and planning performance indicators. The control steady-state performance indicators and vehicle stability indicators are used to evaluate the robot's performance in execution ability; the obstacle avoidance sensitivity index is used to evaluate the robot's performance in judgment ability; the target positioning accuracy index is used to evaluate the robot's performance in perception ability; the planning performance index is used to evaluate the robot's performance in decision-making and planning ability.

[0098] More specifically, the control steady-state performance indicators include average vehicle speed, average left speed, average right speed and average turning angle; the obstacle avoidance sensitivity indicators include average vehicle speed, unit absolute trajectory error and unit relative trajectory error; the vehicle stability indicators include average turning angle; the target positioning accuracy indicators include unit absolute trajectory error and unit relative trajectory error; the planning performance indicators include path time, path distance, planned path smoothness, average turning radius and maximum turning angle.

[0099] Specifically, the indicator system training module is also used for:

[0100] Step 201, constructing a data set;

[0101] Step 202: for the intelligence evaluation, construct an objective function by accumulating various indicators and corresponding indicator data;

[0102] Step 203: Solve the objective function to obtain the weights of various indicators of the evaluation tasks for different types of robots.

[0103] More specifically, step 201 includes:

[0104] First, a specific scenario task is constructed, and a robot of a specified type performs the task multiple times to obtain data indicators of the motion process. Assuming there are n indicators, the task is performed t times, and the tasks performed by the robot are scored at the same time. The data indicators and task scores are combined to obtain a matrix with t rows and (n+1) columns.

[0105] More specifically, step 202 includes:

[0106] The objective function is constructed as F = W1 X1 + W2X2 + ... + W n X n

[0107] Among them, F is the scoring result of task execution, W1,W2,...,W n is the weight of each indicator, which is a variable; X1, X2, ..., X n is the corresponding data indicator;

[0108] The constraints of the objective function are: W1+W2+...+W n =1;W i >=0,W i <=1,i=1,2,...,n.

[0109] More specifically, step 203 includes:

[0110] For different tasks, the ultimate goal of the scoring result F of task execution is to obtain the minimum scoring result or the maximum scoring result. The objective function is solved by the gradient descent method. Through the iterative optimization process, the objective function F is maximized or minimized to obtain the optimal weight, thereby obtaining the weight of each indicator.

[0111] Specifically, the system testing process in the intelligence evaluation module includes:

[0112] Step 301, determine the test object: enter the test platform, and select the robot component or functional module to be tested in the platform;

[0113] Step 302, select intelligent evaluation indicators: select an intelligent evaluation indicator system according to the specific task or scenario of the robot;

[0114] Step 303, adding test tasks and parameters: adding test tasks and setting corresponding test parameters;

[0115] Step 304, obtaining simulation or real data: starting the simulation system for testing, the simulation system performs simulation according to the set test tasks and parameters, and generates simulation data after the simulation is completed or imports motion data collected in the real environment for evaluation;

[0116] Step 305, data processing and evaluation report: obtain simulation or real data and save it in the database, clean and denoise the data, weight the simulation results or real data according to the weights of different indicators to obtain evaluation scores, perform evaluation, and generate an intelligent evaluation report.

[0117] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A robot intelligence evaluation method, characterized in that: The following steps are involved: Step 1: Build an intelligence evaluation index system based on the robot's perception, judgment, decision-making and execution; Step 2: Optimize the training intelligence evaluation indicators and obtain the weights of various indicators for different types of robots; Step 201, constructing a data set; Step 202: for the intelligence evaluation, the objective function is constructed by accumulating each indicator and the corresponding indicator data; the constructed objective function is F = W1 X1 + W2X2 + ... + W n X n ; where F is the score result of task execution, W1, W2, ..., W n is the weight of each indicator, and the weight is a variable; X1, X2, ..., X n is the corresponding data indicator; the constraint condition of the objective function is: W1 + W2+ ... +W n =1; W i >= 0, W i <= 1, i=1, 2, ..., n; Step 203: solving the objective function to obtain the weights of various indicators of the evaluation tasks for different types of robots; Step 3: Build a system testing process and select intelligence evaluation indicators to evaluate the intelligence of the robot.

2. A robot intelligence evaluation method according to claim 1, characterized in that: The step one includes: the intelligence evaluation index system includes a control steady-state performance index, an obstacle avoidance sensitivity index, a vehicle body stability index, a target positioning accuracy index and a planning performance index. The control steady-state performance index and the vehicle body stability index are used to evaluate the robot's performance in execution capability; the obstacle avoidance sensitivity index is used to evaluate the robot's performance in judgment capability; the target positioning accuracy index is used to evaluate the robot's performance in perception capability; and the planning performance index is used to evaluate the robot's performance in decision-making and planning capability.

3. A robot intelligence evaluation method according to claim 2, characterized in that: The control steady-state performance index includes average vehicle speed, average left speed, average right speed and average turning angle; the obstacle avoidance sensitivity index includes average vehicle speed, unit absolute trajectory error and unit relative trajectory error; the vehicle stability index includes average turning angle; The target positioning accuracy indicators include unit absolute trajectory error and unit relative trajectory error; the planning performance indicators include path time, path distance, planned path smoothness, average turning radius and maximum turning angle.

4. A robot intelligence evaluation method according to claim 1, characterized in that: The step 201 includes: First, a specific scenario task is constructed, and a robot of a specified type performs the task multiple times to obtain data indicators of the motion process. Assuming there are n indicators, the task is performed t times, and the tasks performed by the robot are scored at the same time. The data indicators and task scores are combined to obtain a matrix with t rows and (n+1) columns.

5. A robot intelligence evaluation method according to claim 1, characterized in that: The step 203 includes: For different tasks, the ultimate goal of the scoring result F of task execution is to obtain the minimum scoring result or the maximum scoring result. The objective function is solved by the gradient descent method. Through the iterative optimization process, the objective function F is maximized or minimized to obtain the optimal weight, thereby obtaining the weight of each indicator.

6. A robot intelligence evaluation method according to claim 1, characterized in that: The system testing process in step 3 includes: Step 301, determine the test object: enter the test platform, and select the robot component or functional module to be tested in the platform; Step 302, select intelligent evaluation indicators: select an intelligent evaluation indicator system according to the specific task or scenario of the robot; Step 303, adding test tasks and parameters: adding test tasks and setting corresponding test parameters; Step 304, obtaining simulation or real data: starting the simulation system for testing, the simulation system performs simulation according to the set test tasks and parameters, and generates simulation data after the simulation is completed or imports motion data collected in the real environment for evaluation; Step 305, data processing and evaluation report: obtain simulation or real data and save it in the database, clean and denoise the data, weight the simulation results or real data according to the weights of different indicators to obtain evaluation scores, perform evaluation, and generate an intelligent evaluation report.

7. A robot intelligence evaluation system, characterized in that: include: The indicator system building module is used to build an intelligence evaluation indicator system based on the robot's perception, judgment, decision-making and execution; The indicator system training module is used to optimize the training intelligence evaluation indicators and obtain the weights of various indicators for different types of robots; the indicator system training module is also used to: Step 201, constructing a data set; Step 202: for the intelligence evaluation, the objective function is constructed by accumulating each indicator and the corresponding indicator data; the constructed objective function is F = W1 X1 + W2X2 + ... + W n X n ; where F is the score result of task execution, W1, W2, ..., W n is the weight of each indicator, and the weight is a variable; X1, X2, ..., X n is the corresponding data indicator; the constraint condition of the objective function is: W1 + W2+ ... +W n =1; W i >= 0, W i <= 1, i=1, 2, ..., n; Step 203: solving the objective function to obtain the weights of various indicators of the evaluation tasks for different types of robots; The intelligence evaluation module is used to build a system testing process and select intelligence evaluation indicators to evaluate the intelligence of the robot.

8. A robot intelligence evaluation system according to claim 7, characterized in that: The index system construction module is also used for: the intelligence evaluation index system includes control steady-state performance indicators, obstacle avoidance sensitivity indicators, vehicle stability indicators, target positioning accuracy indicators and planning performance indicators. The control steady-state performance indicators and vehicle stability indicators are used to evaluate the robot's performance in execution ability; the obstacle avoidance sensitivity index is used to evaluate the robot's performance in judgment ability; the target positioning accuracy index is used to evaluate the robot's performance in perception ability; the planning performance index is used to evaluate the robot's performance in decision-making and planning ability.

Images